Mailing-List: contact commits-help@airflow.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@airflow.incubator.apache.org
Date: Sat, 28 May 2016 00:46:12 +0000 (UTC)
From: "Lance Norskog (JIRA)" <jira@apache.org>
To: commits@airflow.incubator.apache.org
Message-ID: <JIRA.12965599.1462542958000.322878.1464396372816@Atlassian.JIRA>
In-Reply-To: <JIRA.12965599.1462542958000@Atlassian.JIRA>
References: <JIRA.12965599.1462542958000@Atlassian.JIRA> <JIRA.12965599.1462542958714@arcas>
Subject: [jira] [Commented] (AIRFLOW-58) Add bulk_dump abstract method to
 DbApiHook
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Sat, 28 May 2016 00:46:17 -0000


    [ https://issues.apache.org/jira/browse/AIRFLOW-58?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15305057#comment-15305057 ] 

Lance Norskog commented on AIRFLOW-58:
--------------------------------------

Copying data in bulk turns out to be a really messy problem. It will take a lot of work and air(flow) time before you have a reasonably complete solution. For example, you're going to want to save a bulk copy task as a file and then restart the task in the middle.

We use the Embulk project to do table-to-table copies. We packaged Embulk as a web service and call it from a custom Operator. I can look into open-sourcing our work. Embulk is a Java app, so the web service needs a multi-gig machine to run.
[http://www.embulk.org/docs/]

We would prefer to use the Sqoop program, but that's off-limits since we're converting from Hadoop.

> Add bulk_dump abstract method to DbApiHook
> ------------------------------------------
>
>                 Key: AIRFLOW-58
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-58
>             Project: Apache Airflow
>          Issue Type: Improvement
>          Components: hooks
>    Affects Versions: Airflow 1.7.0
>            Reporter: Bence Nagy
>            Assignee: Bence Nagy
>            Priority: Trivial
>
> I just see no reason for having a method for bulk loading but not for the inverse.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)